The content of the invention
In view of this, the present invention proposes a kind of based on UWB and pedestrian's system for tracking of laser radar mixed positioning and side
Method, meet that the accurate of pedestrian follows and positioned under complex environment, carries out preliminary positioning and identification to pedestrian, so using UWB
Carry out accurate positioning and identification to pedestrian using laser radar afterwards, finally allow the robot to without collision accurately with retinue
People, precision is higher, and stability is stronger.
The technical proposal of the invention is realized in this way:It is fixed based on UWB and laser radar mixing the invention provides one kind
Pedestrian's system for tracking of position, it includes at least three UWB base stations, UWB labels, data processing equipment and motion planning and robot control dress
Put, in addition to laser radar, wherein,
UWB labels, which are placed in, to be followed in target, receives the UWB signal sent from UWB base stations;
UWB base stations send UWB signal, and at least three positions are fixed into the UWB base stations of triangle in robot;
Laser radar, measurement follow the distance between target and robot and angle, robot measurement ambient condition information;
Data processing equipment, by UWB base stations and UWB tag computations follow between target and robot UWB distance and
UWB angles, and the distance between target and robot and angle are followed according to lidar measurement, to UWB distances and UWB angles
Degree is corrected, and generates a reference path;According to the robot information of lidar measurement, in reference path
On the basis of generate a collisionless final path;For final coordinates measurement rate control instruction, and to the speed of generation
Control instruction is filtered processing, obtains smooth rate control instruction;
Robot movement control device, the rate control instruction of data processing equipment processing is converted into movement executing mechanism
Execute instruction.
Second aspect, the invention provides a kind of pedestrian's follower method based on UWB and laser radar mixed positioning, including
Following steps,
S1, UWB signal is sent to UWB labels by UWB base stations, calculates the UWB distances followed between target and robot
And UWB angles;
S2, the distance between target and robot and angle are followed by lidar measurement, to UWB distances and UWB angles
Degree is corrected, it is assumed that clear around robot, generates a nearest reference path;
S3, by lidar measurement robot information, local map is created, then according to the machine after correction
Device people generates a collisionless final path relative to the distance and angle for following target on the basis of reference path;
S4, processing is filtered for final coordinates measurement rate control instruction, and to the rate control instruction of generation, is obtained
To smooth rate control instruction;
S5, the execution that smooth rate control instruction is converted into movement executing mechanism by robot movement control device refer to
Order.
On the basis of above technical scheme, it is preferred that in the step S1, according to principle of triangulation robot measurement
Relative to UWB distance S1 and the UWB angle, θ 1 of UWB labels, position (S1, θ 1) of the UWB labels relative to robot is obtained.
It is further preferred that in the step S2, a frame laser radar data is gathered and then according to UWB by laser radar
The determination of angle, θ 1 follows size (θ 1- Δ/2, θ 1+ Δ/2) of the target in the range of lidar measurement, wherein, Δ is to follow mesh
The measurement range of laser radar corresponding to target width;Then, search follows target phase in the range of (θ 1- Δ/2, θ 1+ Δ/2)
For the position of robot, the point of the n laser radar measured is added, is removed first in this n point in scope (S1- Δs
Noise spot in s, S1+ Δ s), wherein, wherein S1 is UWB distances, and Δ s is the Breadth Maximum of 2 times of human legs, obtains n1 point,
Then the average distance for following target to laser radar is soughtAnd willAs robot and after following target correction
Distance, further calculating robot is with following the angle after target correctionWherein Δ θ is point of laser radar
Resolution, then finally follow position of the target under robot coordinate to be
Still more preferably, in the step S2, reference path was that the origin of robot coordinate system points toPoint
Straight line, it controls linear velocity and the angular speed to beWherein K1 is the ratio pass of distance and linear velocity
System, K2 are Schemes of Angular Velocity Estimation for Robots and robot front and the normal angle proportionate relationship for following target measurement width.
On the basis of above technical scheme, it is preferred that in the step S3, robot is carried out using dynamic window method
Avoidance on travel path, so as to obtain final path, dynamic window method comprises the following steps,
S3-1, laser radar search space is constrained to the controllable control instruction space of robot;
S3-2, the order in search space is evaluated using object function, selection makes object function maximumlly refer to
Order is used as optimum instruction.
It is further preferred that in the step S3-1, robot speed's scope is determined using below equation,
θt=θt+ωΔt
Vm={ v ∈ [vmin,vmax],ω∈[ωmin,ωmax]}
Wherein, r is the radius that robot does circular motion;V is robot linear velocity;ω is robot rotary speed;
Robot coordinate is (x, y), θtIt is robot in the course angle of t, t is current time;
VmFor robot starting velocity space, VdIt is robot up to the velocity space, vc,ωcIt is the current speed of robot
Degree,For the maximum deceleration of robot,For the peak acceleration of robot,For the maximum angular acceleration of robot,
For the maximum angular deceleration of robot, VaThe feasible speed space of barrier is not struck against for robot, dist (v, ω) is speed
The nearest distance of (v, ω) corresponding trajectory distance barrier.
On the basis of above technical scheme, it is preferred that in the step S3-2, every track is entered using below equation
Row evaluation,
Wherein, H (v, ω) is used for evaluating robot under the sample rate currently set, when reaching analog track end
Differential seat angle between direction and target;
G (v, ω) is track evaluation function, and deviation is smaller, and evaluation of estimate is higher, and distance is bigger, and evaluation of estimate is higher, and speed is got over
Greatly, evaluation of estimate is higher;
D (v, ω) represents distance of the robot on current track between nearest barrier;
V (v, ω) is used for evaluating the velocity magnitude of current track, and α, beta, gamma is weight coefficient;
N is all tracks of sampling, and i is current track to be evaluated;
H (i) is differential seat angle of the robot between the direction and target of i-th trailing end away from d (i) is that robot exists
The distance between with nearest barrier on i-th track, v (i) is speed of the robot on i-th track.
On the basis of above technical scheme, it is preferred that in the step S4, the speed using Kalman filter to generation
Control instruction is filtered processing, the signal and the state-space model of noise pre-established, in implementation procedure, according to current moment
Observation and previous moment estimate update the estimation to state variable, reach linear optimal effect.
Pedestrian's system for tracking method based on UWB and laser radar mixed positioning of the present invention has relative to prior art
Following beneficial effect:
(1) preliminary positioning and identification are carried out to pedestrian using UWB, then pedestrian carried out using laser radar accurate
Positioning and identification, effective avoidance, meet that the accurate of pedestrian follows and positioned under complex environment;
(2) avoidance is carried out using dynamic window method, precision is higher;
(3) processing is filtered using motion of the Kalman filter to robot, stability is stronger.
Embodiment
Below in conjunction with the accompanying drawing in embodiment of the present invention, the technical scheme in embodiment of the present invention is carried out clear
Chu, it is fully described by, it is clear that described embodiment only a part of embodiment of the present invention, rather than whole realities
Apply mode.Based on the embodiment in the present invention, those of ordinary skill in the art institute under the premise of creative work is not made
The every other embodiment obtained, belongs to the scope of protection of the invention.
As shown in figure 1, the UWB of the present invention and pedestrian's system for tracking of laser radar mixed positioning, it includes laser radar
1st, at least three UWB base stations 2, UWB labels 3, data processing equipment 4 and robot movement control device 5.
UWB labels 3, which are placed in, to be followed in target, receives the UWB signal sent from UWB base stations 2;
UWB base stations 2 send UWB signal, and at least three positions are fixed into the UWB base stations 2 of triangle in robot;
Laser radar 1, measurement follow the distance between target and robot and angle, robot measurement surrounding environment letter
Breath;
Data processing equipment 4, the UWB distances followed between target and robot are calculated by UWB base stations 2 and UWB labels 3
And UWB angles, and according to laser radar 1 measure follow the distance between target and robot and angle, to UWB apart from and
UWB angles are corrected, and generate a reference path;The robot information measured according to laser radar 1, is joining
Examine and a collisionless final path is generated on the basis of path;For final coordinates measurement rate control instruction, and to generation
Rate control instruction be filtered processing, obtain smooth rate control instruction;
Robot movement control device 5, the rate control instruction that data processing equipment 4 is handled is converted into Motor execution machine
The execute instruction of structure.
Pedestrian's follower method based on UWB and laser radar mixed positioning of the present invention, comprises the following steps,
S1, UWB signal is sent to UWB labels 3 by UWB base stations 2, calculate follow UWB between target and robot away from
From and UWB angles.
Specifically, according to principle of triangulation robot measurement relative to UWB distance S1 and the UWB angle, θs of UWB labels 3
1, obtain position (S1, θ 1) of the UWB labels 3 relative to robot.
Principle of triangulation:Two-way time-of-flight method, the arteries and veins of Ta1 transmitting request property of the UWB base stations 2 on its timestamp
Signal is rushed, by UWB labels 3 in Ta2 receptions, UWB labels 3 launch the signal of a response property at the Tb1 moment, by UWB
Time stamp T b2 reception of the base station 2 at oneself.Pulse signal can be calculated successively between UWB base stations 2 and UWB labels 3
Flight time, so that it is determined that flying distance S.S=C × [(Ta2-Ta1)-(Tb2-Tb1)] (C is the light velocity), then by
Multiple cans of diverse location arrangement UWB base stations 2 measure UWB labels 3 relative to machine by principle of triangulation in robot
The position (S1, θ 1) of device people.
S2, the distance between target and robot and angle are followed by the measurement of laser radar 1, to UWB distances and UWB angles
Degree is corrected, it is assumed that clear around robot, generates a nearest reference path.
One frame laser radar data is gathered by laser radar 1 and then is determined to follow target in lidar measurement according to θ 1
In the range of size (θ 1- Δ/2, θ 1+ Δ/2), wherein, Δ is the measurement model for following laser radar 1 corresponding to the width of target
Enclose;Then search follows target to add the n measured relative to the position of robot in the range of (θ 1- Δ/2, θ 1+ Δ/2)
The point of individual laser radar 1, removed first in this n point scope (noise spot in S1- Δs s, S1+ Δ s), wherein, wherein
S1 is UWB distances, and Δ s is the Breadth Maximum of 2 times of human legs, obtains n1 point, then asks and follows target to the flat of laser radar 1
Equal distanceAnd willAs robot and the distance after target correction is followed, further calculating robot and follows mesh
Angle after calibration justWherein Δ θ is the resolution ratio of laser radar 1, then finally follows target in robot
Position under coordinate is
Reference path was that the origin of robot coordinate system points toThe straight line of point, it controls linear velocity and angular speed
ForWherein K1 is the proportionate relationship of distance and linear velocity, and K2 is Schemes of Angular Velocity Estimation for Robots and robot
Front and the normal angle proportionate relationship for following target measurement width.
S3, by the robot measurement ambient condition information of laser radar 1, create local map, then according to correction after
Robot generates a collisionless final path relative to the distance and angle for following target on the basis of reference path.
Robot is tracking target, it is possible that the barrier incoherent with target, can influence robotic tracking
Target, the present invention carry out avoidance using dynamic window method (dwa), and dynamic window method is directly to be searched in control instruction space
The automatic obstacle avoiding algorithm for making object function take the Optimal Control of maximum to instruct.This method can be summarized as two steps:First, by laser
The search space of radar 1 is constrained to the controllable control instruction space of robot;Second, using object function to the life in search space
Order is evaluated, and selection makes object function maximumlly instruct as optimum instruction.Concrete principle is as follows:
Assuming that robot is not omnidirectional moving robot, what rotation of advancing can only be carried out, a pair (v, ω) just represent one
Arc track, the radius that it does circular motion be,
Wherein, r is the radius that robot does circular motion;V is robot linear velocity;ω is robot rotary speed.
When rotary speed ω is not equal to 0, robot coordinate is,
θtt=θt+ωΔt
Wherein, robot coordinate is (x, y), θtIt is robot in the course angle of t, t is current time;
Track is extrapolated according to speed can, then evaluates these tracks.
Limitation of the mobile robot by the kinematical constraint of itself, i.e. maximal rate and minimum speed:
Vm={ v ∈ [vmin,vmax],ω∈[ωmin,ωmax]}
Wherein, VmFor robot starting velocity space.
Because the motor torque of robot is limited, the acceleration and deceleration limitation of maximum be present, therefore mobile robot is in a meter
Calculate in the cycle, a dynamic window be present, the speed in the window is the speed that robot can actually reach:
VdIt is robot up to the velocity space, vc,ωcIt is the present speed of robot,For the maximum deceleration of robot
Degree,For the peak acceleration of robot,For the maximum angular acceleration of robot,For the maximum angular deceleration of robot;
Further contemplating robot be able to can stop before barrier is encountered, therefore under the conditions of maximum deceleration, speed has
One scope:
Wherein, VaThe feasible speed space of barrier is not struck against for robot, dist (v, ω) is corresponding for speed (v, ω)
The nearest distance of trajectory distance barrier.
In the velocity group of sampling, it is feasible to have some groups of tracks, therefore uses the mode of evaluation function as every rail
Mark is evaluated.Evaluation function is:
H (v, ω) is for evaluating robot under the sample rate currently set, reaching court during analog track end
To the differential seat angle between target;
G (v, ω) is track evaluation function, and deviation is smaller, and evaluation of estimate is higher, and distance is bigger, and evaluation of estimate is higher, and speed is got over
Greatly, evaluation of estimate is higher;
D (v, ω) represents distance of the robot on current track between nearest barrier, if on this track
There is no barrier, that just sets it to a constant;
V (v, ω) is used for evaluating the velocity magnitude of current track, and α, beta, gamma is weight coefficient;
In order to avoid occur a certain item in evaluation function it is too dominant, place is normalized to each single item of evaluation function
Reason:
Wherein, n is all tracks of sampling, and i is current track to be evaluated;
H (i) is differential seat angle of the robot between the direction and target of i-th trailing end away from d (i) is that robot exists
The distance between with nearest barrier on i-th track, v (i) is speed of the robot on i-th track.
In each calculating cycle, the obstacle information and target information around robot are obtained by laser radar 1, obtained
To feasible track, and by calculating the value of evaluation function corresponding to every track, evaluation function is set to obtain maximum corresponding
The set-point of (v, ω) next period velocity and angular speed corresponding to track.
S4, processing is filtered for final coordinates measurement rate control instruction, and to the rate control instruction of generation, is obtained
To smooth rate control instruction.
When tracking target person motion, jitter phenomenon affected by noise can occur in speed, while also have and measure for robot
Error, robot motion is filtered used here as Kalman filter.Kalman filter is built in advance using recursive algorithm
Vertical signal and the state-space model of noise, in implementation procedure, according to the observation of current moment and previous moment estimate
To update the estimation to state variable, reach linear optimal effect.The algorithm is concise, and has only used previous moment
Data, so Kalman filtering is adapted to and computer real-time operation.Filtering principle is as follows:
The motion state equation of system
X (k)=AX (k-1)+Wk
Wherein, Wk is process noise amount, and such a noise is white Gaussian noise, meets wk~(0, Q (k)).
X (k)=[lk,vl.k,θk,vθ.k]TState variable is respectively to follow the length at target k moment, linear velocity, angle, angle
Speed.
State-transition matrix:For the controlling cycle of system.
The observational equation of system is:
Z (k)=HX (k)+υk
Wherein, υkIt is the noise vector of observation, such a noise is that white Gaussian noise meets υk~(0, R (k)).
Observing matrix:
If the state X (k-1 | k-1) of the system at our known system k-1 moment and its corresponding covariance P (k-1 | k-
1) the state X (k | k-1) and its corresponding covariance P (k | k-1) of the system of subsequent time can, be predicted
Predictive equation is
X (k | k-1)=AX (k-1 | k-1)
P (k | k-1)=AP (k-1 | k-1) AT+Q (1)
Then we can obtain the observation Z (k) at k moment, and by predicted value X (k | k-1) and its corresponding covariance P
(k | k-1) obtains X (k | k) and its corresponding covariance P (k | k).Wherein X (k | k) is the estimate to system actual value.
Make kalman gain
Kg (k)=P (k | k-1) HT(HP(k|k-1)HT+R)-1 (2)
It can so update X (k | k), and P (k | k), renewal equation is as follows:
X (k | k)=c+Kg (k) [Z (k)-HX (k | k-1)]
P (k | k)=[I-Kg (k) H] P (k | k-1) (3)
Each k moment we by using human body measurement position Z (k) and (2) (3) update Karman equation, and
The estimate X (k | k) of position of human body is obtained, predicted value X (k+1 | k) is then obtained according to formula (1), and put using this and be used as prediction
Value.Such iterative cycles reach filter effect.
S5, the execution that smooth rate control instruction is converted into movement executing mechanism by robot movement control device 5 refer to
Order.
The better embodiment of the present invention is the foregoing is only, is not intended to limit the invention, it is all the present invention's
Within spirit and principle, any modification, equivalent substitution and improvements made etc., it should be included in the scope of the protection.